Collaboration with YouTube

Applying our AI research to improve the YouTube experience

To enrich people’s lives through our research, we’ve partnered with Alphabet companies to apply our technology to improve the products and services used by billions of people every day.

One of our main partners is YouTube, which is on a mission to give everyone a voice and show them the world.

Working with YouTube’s product and engineering teams, we’ve helped optimize decision-making processes that increase security, reduce latency, and improve the experience for viewers, creators, and advertisers for everyone.

Video compression optimization

With the surge in video during the COVID-19 pandemic and the overall amount of internet traffic expected to increase in the future, video compression is an increasingly important issue.

In collaboration with YouTube, we explored the potential of our AI model, MuZero, to improve the VP9 codec, an encoding format that allows videos to be compressed and transmitted over the Internet. Next, we applied MuZero to some of YouTube’s live traffic.

Since launching some of YouTube’s live traffic into production, we’ve demonstrated an average bitrate reduction of 4% across a large and diverse set of videos. Bitrate helps determine the compute capacity and bandwidth needed to play and store videos, which impacts everything from a video’s loading time to its resolution, buffering, and performance. use of data.

By improving the VP9 codec on YouTube, we’ve helped reduce internet traffic, data usage, and the time it takes to load videos. And thanks to optimized video compression, millions of people around the world can watch more videos while using less data.

Protect brand safety for creators and advertisers

Since 2018, we’ve been working with YouTube to better educate creators about the types of videos that can generate revenue from ads and to ensure ads appear alongside content that meets YouTube’s Advertiser-Friendly Guidelines.

Together with the YouTube team, we’ve developed a Labeling Quality Model (LQM) that helps label videos more accurately, in line with YouTube’s friendly ad guidelines. The model improved the accuracy of ads served on videos in accordance with YouTube’s ad adaptation policies.

By improving the way videos are identified and categorized, we’ve built trust in the platform for viewers, creators and advertisers.

Improve Auto Chapters

In recent years, creators have started adding chapters to their videos to make it easier for their audience to find the content they’re looking for, but this manual process can be slow and laborious.

To improve the experience for creators and viewers, we collaborated with the YouTube research team and developed an artificial intelligence system capable of automatically processing video transcripts, audio and visual features and suggesting chapter segments and titles to YouTube creators.

As Sundar Pichai showcased at Google I/O 2022, auto-generated chapters are already available for 8 million videos today, and we plan to expand this feature to over 80 million auto-generated chapters over the course of next year.

With automatic chapters, viewers spend less time searching for specific content and creators save time creating chapters for their videos.

Evolving technologies and products

As society and the technology we use evolves, we are always looking for new ways to improve Alphabet technologies and products every day through our AI research.

Our work with YouTube has already had a big impact, and we hope to bring many more meaningful improvements to people’s lives through our continued collaborations.

Comments are closed.